Optimizing TSES method based on the environmental factors to select negative samples and its application in landslide susceptibility evaluation

Abstract

Objective Landslide susceptibility evaluation is an important means for landslide disaster prevention and control. Unreasonable negative landslide samples will affect landslide susceptibility evaluation, thereby affecting landslide disaster prevention and control. Therefore, it is particularly critical to provide a reasonable negative sample selection method. Methods In view of the selection of negative landslide samples, the ancient landslides in Milin City, Xizang are taken as an example, and 10 environmental factors, including elevation, slope, slope aspect, slope position, distance to road, distance to fault, distance to water system, topographic relief, lithology and land use type, are selected. The Relief algorithm is used to calculate the contribution values of environmental factors and to select the optimized environmental factors. The target space exteriorization sampling (TSES) method based on the optimization of environmental factors is applied to select negative samples as the input variables of the random forest (RF)model with excellent performance.Then, the optimized environmental factors and positive/negative samples are combined to predict the landslide susceptibility of Milin city, and the confusion matrix and receiver operating characteristic (ROC) curve are used to evaluate the prediction performance.To test the effectiveness and advancement of the TSES method optimized by environmental factors, the coupled information method and TSES method are respectively used to select negative landslide samples and constructs RF models to conduct comparative research with the RF model constructed by the TSES method optimized by environmental factors. Results The results show that the evaluation effect of the RF model constructed by the optimized TSES method based on environmental factors is better with an ACC value of 93.7% and an AUC value of 0.987, both of which are greater than those of the RF models constructed by the coupling information method and the TSES method. Conclusion The TSES optimized by environmental factors can improve the accuracy of the RF model, solve the problems of environmental factor selection in multifactor constrained sampling, and provide a new approach to collect negative landslide samples for landslide susceptibility evaluation.

Keywords

• landslide susceptibility evaluation
• relief algorithm
• negative sample
• tses method based on the optimixation of environmental factor
• random forest
• ancient landslide